Entwurf und Realisierung einer flexiblen verteilten Kommunikation fuer einen parallelen evolutionaeren Algorithmus

This report focuses on the parallelization of the evolutionary tools being integrated in the design optimization tool environment SIMOT (SIMulation and Optimization Tool Environment). These optimization tools called GAMA (Genetic Algorithm for Model Adaptation) and GADO (Genetic Algorithm for Design Optimization) have very long run times due the need of time consuming simulators like FEM-simulator, network simulator and optic simulator. The specific simulator calculates the quality of a given model or design which is then used for guiding the evolutionary search process. The as a matter of principle parallel nature of evolutionary algorithms on one hand and the existance of computing power in our network of workstations on the other hand are the pre-conditions for the design and implementation of a parallel version of the evolutionary optimization tools. This thesis gives a short introduction into evolutionary algorithms and describes the various possibilities for a parallelization. The realization of the parallel concept is based on PVM (Parallel Virtual Machine) a well accepted and portable public domain software allowing a heterogeneous collection of workstations and supercomputers to function as a single high-performance parallel machine. (orig.)Diplomarbeit von P. Biegler, Univ. Karlsruhe (DE). Fak. fuer InformatikSIGLEAvailable from TIB Hannover: ZA 5141(6045) / FIZ - Fachinformationszzentrum Karlsruhe / TIB - Technische InformationsbibliothekDEGerman